next up previous
Next: Bibliography Up: Statistical moments - An Previous: Moment noise sensitivity

Conclusions

This section has detailed the conventional use of statistical moments - the analysis of single two-dimensional images. Non-orthogonal and orthogonal descriptors have been discussed, describing moments which possess various useful properties including translation, scale and rotation invariance. We have considered both image description and reconstruction. These techniques are applied to single images and describe a shape in terms of its spatial (or pixel) distribution. However, many computer vision and image processing problems involve the analysis of image sequences. For example, analysing the movement of an aircraft in the sky, or ultra-sound images of a beating heart. These image sequences can be of a rigid shape (i.e. an aircraft) or a deforming one (i.e. a beating heart), while consecutive images within the sequence tend to be highly correlated. Extending moments to the analysis of image sequences has been achieved using a general framework called velocity moments [14,15] - enabling the analysis of moving features within image sequences.



Jamie Shutler 2002-08-15